svn commit: r272393 - head/etc

[Hiroki Sato]

Colin Percival <cperciva at freebsd.org> wrote
  in <542CAABB.5090900 at freebsd.org>:

cp> On 10/01/14 18:16, Hiroki Sato wrote:
cp> >   This is an attempt to solve a problem that rc.d scripts from third-party
cp> >   software do not have entries in /etc/defaults/rc.conf.
cp>
cp> Isn't this why we have the
cp> > : ${foo_enable="NO"}
cp> idiom in ports rc.d files?
cp>
cp> >   The fact that
cp> >   load_rc_config() reads rc.conf only once and /etc/rc invokes the function
cp> >   before running rc.d scripts made developers confused for a long time because
cp> >   load_rc_config() just before run_rc_command() in each rc.d script overrides
cp> >   variables only when the script is directly invoked, not from /etc/rc.
cp>
cp> If a script is setting variables for its own use, there's no need to use
cp> functions from rc.subr -- it can just set the variables directly.  If a
cp> script is editing rc.conf, sending a SIGALRM to $$ will signal /etc/rc to
cp> re-source rc.conf.
cp>
cp> I'm really not clear on what this commit accomplishes.

 The primary purpose is to make it clear which variables are used in
 the script for *user configuration* and provide a consistent writing
 style for scripts from both base and ports.  More specifically, I
 want to implement a way to list user-configurable variables and which
 one is changed from the default value, by effectively replacing
 functionality of defaults/rc.conf with set_rcvar().

 Use of : ${foo="NO"} idiom after load_rc_config() works as you
 pointed out.  However, it does not work with listing variables.
 Plus, there are many scripts written in an inconsistent way.  Some
 scripts call load_rc_config() multiple times, some have the idioms
 between load_rc_config() and run_rc_command(), and some have them
 mistakenly before load_rc_config().  I think this is due to confusion
 about how load_rc_config() works and all of them can be
 fixed/rewritten consistently of course, but I think gathering
 definitions at the head of the scripts and making them being defined
 at the end of load_rc_config() as set_rcvar_obsolete() does are more
 intuitive.

-- Hiroki
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 181 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/svn-src-head/attachments/20141002/3ae9074b/attachment.sig>